Configurable process models

Abstract

Business process models and workow systems aim at guaranteeing efficient and reliable executions of business processes. For this, they require detailed specifications of the individual steps that need to be taken during the process execution. As defining such process definitions from scratch requires good knowledge of both the domain of the process as well as the process modeling technique, process modeling is often cumbersome and prone to errors that inhibit sound process executions. For that reason, many providers of process modeling software as well as process consultants offer so-called reference process models or template repositories. These repositories provide established process specifications for both general as well as industry specific processes, from travel approvals to invoice verifications. Still, even for such common processes, execution variations exist among organizations. Hence, organizations have to adapt the templates manually to individual requirements. The process modeling skills required for these adaptations are equal to the skills required for modeling from scratch. This PhD thesis suggests reducing the need for manual process model adaptations by integrating variations among different process executions into a single process model | a configurable process model. In the configurable model, process configuration allows inhibiting executing any undesired tasks. In this way, a process model providing exactly the behavior desired by an organization can be derived without manual process modeling efforts. For defining configurable process models, the thesis first provides a formal specification of process model configuration by analyzing how behavior is added to process models and defining process configuration as the inverse. In this way, blocking and hiding of behavior, originally defined as concepts to discover inheritance relations among dynamic behavior, are identified as the two techniques to restrict behavior of configurable process models. The practical feasibility of the suggested approach is demonstrated through depicting how process configuration can be added to the process modeling languages of existing workow systems, namely SAP WebFlow, BPEL, and YAWL. An actual implementation is provided for YAWL. It even allows configuring the process model through natural language questions by mapping the various configuration options to pre-defined answers in a questionnaire. The framework has been tested in a case study where configurable process models were developed for registration processes executed in municipalities like the registration of a newborn child or a marriage. The developed models thus allow deriving a configured, individual process variant by simply answering a questionnaire on desired and undesired options of the process. The resulting model variant is directly executable in the YAWL system environment. The models and the overall approach have afterwards also been evaluated through interviews with various stakeholders in the process configuration lifecycle. Two main challenges arise for defining and using configurable process models: On the one hand, the creation of a process model integrating various process variants obviously requires far more work than building a process model covering only one of these variants. For that reason, the thesis suggests a set of process mining techniques, which can help in the construction of a configurable process model, e.g. by merging individual process models. On the other hand, process configuration allows restricting the process behavior depicted in the configurable process model further than desired, up to the point that the process is not executable anymore. This is addressed through discussing a range of constraints on the process model configuration that are able to guarantee sound processes.